With over 70% of the world’s entire population expected to be living in cities by 2050, supporting citizens’ mobility within the urban environment is a priority for municipalities worldwide. Although public multi-modal transit systems are necessary to better manage mobility, they are not sufficient. Citizens must be offered personalized travel information to make their journeys more efficient and enjoyable. Notably, such information should not only be objective (e.g., bus timetable, live bus tracking), but crucially personalized – since every passenger preferences and interests differ (e.g., crowdedness of trains, heat of tube platforms, sociability of the coaches).
To enable this, a multitude of research problems need to be solved. On the one hand, efficient techniques for mobile participatory sensing are required to create robust mobile distributed systems that can provide on-demand sensing information at a large scale. This needs to then be complemented by domain-specific machine learning algorithms, which must be able to execute on resource constrained mobile devices with heterogeneous configurations.
Development of a scalable data collection middleware and personalized mobility services, that gathers both passive sensory information (e.g., bus location from GPS) and active user-generated content (e.g., road hazard reports, crowdedness of journey, etc.), and disseminates transport network status updates to travelers as per their preferences about various aspects of people mobility (e.g., punctuality, crowdedness, and sociability of transport systems). Challenges here include the large scale, system dynamicity, and heterogeneity while combining the data streams of interest to a user.
Learning and mining techniques for sustainability, that can help in identifying events of interest (e.g. road conditions, driving anomalies, user’s social context) from collected data; and help in providing personalized services to complement the first aim. Recently, machine learning has been increasingly used in enhancing capabilities of smartphones. It has been used in detecting everyday activities, road conditions, driving behavior, inferring information from complex sensors like audio etc. We will study different applications of machine learning in providing personalized services and develop new approaches that enable such services in urban transport context.
Development of a demonstrator for real-life assessment, and deployment with actual end-users.